Infection-related preterm birth (PTB) is more common at early gestational ages and is associated with major neonatal mortality and morbidity. Abnormal genital tract microflora in early pregnancy predicts late miscarriage and early PTB. Accordingly, it is logical to consider antibiotics as an intervention. Unfortunately, the conclusions of systematic reviews and meta-analyses (SR&MAs) carried out in an attempt to explain the confusion over the heterogeneity of individual studies are flawed by the fact that undue reliance was placed on studies which: (a) had a suboptimal choice of antibiotic (mainly metronidazole) or used antibiotics not recommended for the treatment of bacterial vaginosis (BV) or BV-related organisms; (b) used antibiotics too late in pregnancy to influence outcome (23–27 weeks); and (c) included women whose risk of PTB was not due to abnormal genital tract colonization and hence unlikely to respond to antibiotics. These risks included: (a) previous PTB of indeterminate etiology; (b) low weight/body mass index; or (c) detection of fetal fibronectin, ureaplasmas, Group B streptococcus or Trichomonas vaginalis). While individual studies have found benefit of antibiotic intervention for the prevention of PTB, in meta-analyses these effects have been negated by large methodologically flawed studies with negative results. As a result, many clinicians think that any antibiotic given at any time in pregnancy to any woman at risk of PTB will cause more harm than good. Recently, a more focused SR&MA has demonstrated that antibiotics active against BV-related organisms, used in women whose risk of PTB is due to abnormal microflora, and used early in pregnancy before irreversible inflammatory damage has occurred, can reduce the rate of PTB. This review presents those data, the background and attempts to explain the confusion using new information from culture-independent molecular-based techniques. It also gives guidance on the structure of putative future antibiotic intervention studies.